Convolutional Neural Network Based Fruit Image Classification

Abstract

Automated categorization of freshness of fruits plays a pivotal role in the agricultural industry. In conventional methods, the grading of fruit is assessed by human beings. Traditional fruit classification methods have often relied on manual operations based on visual ability, and such methods are tedious, time-consuming, and inconsistent. Therefore, a fast, accurate, and automated system is required for industrial applications. This paper presents a Convolutional Neural Network (CNN) based Fruit Image Classification model. The proposed CNN model is implemented using a public dataset named "fruit fresh and rotten for classification" derived from Kaggle. Using the dataset, varieties of fresh fruits (Apple, Banana, and Oranges) and their rotten categories are used for experimentation. Before modeling, the dataset is divided such that 80% is used for training and the remaining 20% for testing. Results show that the proposed CNN model works efficiently in classifying fruits. The accuracy rate (98.23%) of the proposed approach is better than SVM and NB models.